Skip to content
Apart Research
Apart Research
Contact us
Sprint projectOct 25, 2026Hong Kong

Managing Collusion in AI Agents

Ambrose Chan, William Yuen, Michael Choi, Andy Chan, Michael Kit · Team Dragon HK

Submitted to AI Collusion Research Sprint. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Managing Collusion in AI Agents

Open in a new tab (opens in new tab)
Share

Autonomous agents can sustain supra-competitive outcomes without explicit agreement. We ask which variables are necessary for collusion, whether information-theoretic statistics can detect it, and which single restriction removes the most collusion at the least cost. In a Calvano-style Bertrand market with independent Q-learners, full observation of rival prices is sufficient for stable positive markups in two-agent markets (up to +0.22). Sequential timing, cheap talk, and side-payments are not necessary once that information is present. With three agents, a compact summary of rival prices restores collusion. A pre-registered comparison separates absolute suppression from cost-adjusted performance: anti-collusion penalties remove the most markup; hiding prices or small price noise is the least-cost way to return average markup near Nash. Residual mutual information does not detect collusion on learned paths (AUROC 0.35-0.65). In a complementary marketplace laboratory, an elicitation ladder produces capture only at an explicit hint, and a trusted-agent pilot shows that speech and action can dissociate. Both rankings, and the negative detection result, are reported.

Reviews

Judging this Sprint?

Review “Managing Collusion in AI Agents”

Your public critique appears on the project page without your name. Reviews of strong projects are shared with funders, as you choose below.

Not shown on the project page.

Shown on the project page, without your name.

Only the Apart team reads this, and funders if you agree below.

Decline to score

Only when the quality of the writing makes the work impossible to follow. Say why in your private critique; the project page shows only that a judge declined, not why.

How much would this matter for the field if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.
Sharing your name *

When we share a strong project with funders, your name, scores and public critique go with it, confidentially; this choice is only about whether your name is public. If you think this project deserves funding, please also endorse it: that carries the most weight.

Share my private critique with funders *

See our Privacy Policy.

  1. Separating the strongest restriction from the cheapest one is a useful framing, and the limitations section is honest. However, only 2 of 5 baseline seeds collude, so the ranking of interventions is hard to trust. Three of the six pre-registered arms (fine, escrow and the penalty on both agents) are missing from the results. The Track B finding that what agents say and do can diverge is interesting, but it rests on a single run with an unnamed LLM

Cite this project

@misc{chan2026managing,
  title = {{Managing Collusion in AI Agents}},
  author = {Ambrose Chan and William Yuen and Michael Choi and Andy Chan and Michael Kit},
  year = {2026},
  month = oct,
  note = {Submitted to AI Collusion Research Sprint, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/managing-collusion-in-ai-agents-bc1r}},
  url = {https://apartresearch.com/sprints/projects/managing-collusion-in-ai-agents-bc1r}
}

More from AI Collusion Research Sprint

All 392 entriesBrowse all projects

A Sprint is running now

AI x Epistemics Research Sprint · Nov 13 - 15, 2026